That’s why he’s designed his END450 capstone seminar for seniors in the BA in Environmental Design program with “AI as a mindset” – a concept borrowed from his co-instructor Mo Han, principal research scientist at Autodesk Research – as students develop proposals for new spaces of social infrastructure and community connection in downtown Buffalo.

“People have different philosophical propositions towards AI. My proposition is that it's already here; it's irresistible. It's like a flood, with two ways to approach it: you can clog the waterway to stop the flood, or you can direct the water into other areas.”

“I'm more of, you know, a water guy. Let the water flow. So, if it's out there and people are using it, how can they use it better, more effectively, and more responsibly? That's the question,” says Nie, who joined UB in 2023 with undergraduate and graduate degrees in architecture and urban design, including a Master of Architecture and Doctor of Design from the Harvard Graduate School of Design, along with several years of experiences as a practitioner in Boston, New York, and Shanghai.

His research, which evolved through his doctorate studies and a post-doctoral fellowship with the Harvard T.H. Chan School of Public Health, examines how infrastructures of urban social reproduction shape urban inequality, governance, and everyday life across the United States, China, and Southeast Asia. Specifically, he investigates healthcare and social infrastructures, such as hospitals, civic spaces, and participatory platforms, that organize access to resources and opportunities, structure spatial hierarchies, and mediate inclusive and exclusive practices.

Nie employs AI in his own research and received a 2025 award from the UB Baldy Center for Law and Social Policy to explore the potential of AI for analyzing planning regulatory frameworks. He was the primary author of a recently published article in the Journal of Planning Education and Research that deciphered patterns of citizen influence in local planning decisions through a large-language-model analysis of more than 6,400 public meeting transcripts.

Indeed, higher education is increasingly embracing AI rather than resisting it. With rapid adoption across industries and disciplines – and the introduction of new, higher-powered tools almost daily – students need to be better equipped to scrutinize AI outputs, recognize errors or bias, and responsibly leverage its power.

In the space of urban planning and environmental design, “Urban AI” is transforming the way the industry understands, designs, and governs urban systems. Proliferating technologies like machine learning, remote sensing, and computer vision now discern patterns in datasets as diverse as zoning codes, street-level imagery, and wearable technology outputs. AI tools can map heat resilience down to the level of households, optimize traffic and mobility, simulate urban development impacts, measure neighborhood walkability, and accelerate the analysis of large volumes of textual and spatial data. Yet human judgement and expertise remain essential to designing prompts, detecting AI “hallucinations,” accounting for algorithmic bias, and, perhaps most importantly, balancing the complex, contested, and place-based values of planning.

The Association of Collegiate Schools of Planning’s Urban Planning AI Task Force, which released its final report in September, advises that AI’s public diffusion and reduced technical barriers make its integration into urban planning education inevitable. “AI adoption can now occur…without an explicit institutional decision because AI features are increasingly embedded in productivity software, GIS platforms, search systems, cloud services, public information tools, and even word-processing tools. The distinction between choosing to adopt AI and being enrolled in AI-enabled systems is therefore becoming increasingly difficult to maintain,” the report states.

With an affinity for metaphors, Nie likens AI to a “sky magnet,” with the ability to pull students up to great heights, but without the scaffolding underneath, leaving precipitous gaps in critical thinking. That’s where he – and courses like END450 – come in.

“AI tools can deliver superb results. But that’s not the purpose of education. Education is about developing a way of thinking,” says Nie. “AI is an amplifier, but better thinkers get better amplification. The not-so-good thinkers tend to get stuck.”

“The goal is not simply to produce work with AI, but to understand when AI is useful, when it is unreliable, and how its outputs can be checked, interpreted, and turned into defensible planning and design solutions,” Nie continues.

Nie launched the studio last year as a pilot, guiding students along a three-phased pathway of AI-embedded assignments, from argument formation through planning document analysis, to “reality testing” against current policy and spatial conditions, to proposal development and implementation planning. But he observed wide gaps in students’ AI literacy, particularly when it came to breaking down large tasks, processing data about urban life, and verifying complex dynamics like stakeholder networks.

This year, he’s divided the course into eight smaller, scaffolded assignments. Each integrates analog and digital research and is backed by an in-class demonstration of the AI-assistive tool, with detailed requirements for AI chat logs, output verification, and evidence of discernment. “Anything produced by AI is just a claim,” cautions Nie, who combines his teaching with regular lectures from Mo Han.

The studio and its AI focus shine a spotlight on a critical district of Buffalo. “Downtown Buffalo has always been a struggle for the city, and public use of its social spaces has decreased, particularly since the COVID pandemic. At the same time, new investment and public conversations about its future have created opportunities to rethink downtown as a place of connection.," says Nie. 

Students started the semester with a literature review employing Retrieval-Augmented Generation (RAG), a method that allows a language model to generate answers using a set of designated documents, to locate relevant passages, explain discrepancies between authors, and organize relationships across the readings. “AI can help you navigate the literature. It cannot decide what your argument is,” the syllabus states.

Next, RAG is used to read Buffalo’s “Green Code,” a 300-plus-page document riddled with technical language, combined with site visits and Google Street View images, for a comparison of allowable and existing land uses along Main Street, downtown Buffalo’s social and commercial artery. “The gap between what is allowed and what is already there shows you the potential for development,” Nie states. Students must document their RAG setup, log their questions, and verify every AI answer with code citations.

Their investigations continue with a multi-year analysis of location data from more than 32 million mobile phone users across major cities in the United States, including Buffalo, drawn from a publicly available dataset published in Scientific Data by Yan Wang and Zhan Guo of the Florida Institute for Built Environment Resilience at the University of Florida. “The goal is to decide what variables make the most sense, find patterns in Buffalo….and conclude with an argument about what Buffalo’s use of social infrastructure looks like across a year, and what it would take to reshape it.”

“It’s deceiving,” Nie reflects. “AI will fabricate a variable on the Y axis if the prompt or the student’s thinking wasn’t clear. So, I’ll ask my students, what does the Y axis mean? Which column in the dataset did you use? If they don’t know, then they have to go back to work.”

Next, they test the feasibility of their proposals against 88 best practices in American cities catalogued by the UB Rudy Bruner Center for Urban Excellence, in partnership with Lauren Fischer, the center’s director of research. Students examine project narratives – from outcomes to stakeholder considerations – to form a case for three “moves” they would transfer to their own proposal. They must document why they chose those moves over other alternatives, note what the AI model got wrong (and how they caught it), check all financing figures, and watch for data hallucinations.

Finally, students shift into proposal development, forming groups of three around similar ideas. AI is used to synthesize their analyses, compare individual student proposals, test program combinations, and identify questions the group needs to resolve. AI-assisted visualization tools render their ideas into plans, section drawings that show spatial relationships among program elements, and diagrams for ideas that can’t be visualized. “Your visualizations should represent your design, not redesign it by accident,” Nie advises his students.

AI helps structure an implementation plan by organizing potential stakeholders, funding sources, and regulatory pathways. Each claim must be explored and verified, as AI has the potential to miss recent changes in funding program requirements or nuanced alignment of organizational missions.

In December, students will defend their proposals to a jury of Autodesk representatives and researchers from the Rudy Bruner Center for Urban Excellence, which is advancing several community-based planning and design projects focused on downtown Buffalo. Students will have prepared their presentations with AI as a critic by proxy, inviting it to test talking points for different audiences and challenge their assumptions, but being careful not to let it introduce new claims “just because they improve the story,” says Nie.

“AI can bolster our capacity, but with that we lose autonomy,” he continues. “The best way to get that autonomy back is to break things down into smaller tasks. It’s a workflow and way of thinking I’m trying to teach them that hopefully they can incorporate into future projects and use as practicing planners.”

“By developing responsible use of AI, students are also learning to be responsible for themselves,” reflects Nie, who plans to expand his END450 pilot to other areas of Buffalo and use the course to inform departmental dialogue on AI-integrated teaching.